{"paper":{"title":"Recursive Self-Improvement through Multi-Agent Self-Supervision","arxiv_id":"2610.12176","primary_cat":"cs.AI","submitted_at":"2026-10-08T15:44:02+00:00","version_compiled":"v1"},"notice":"Every quoted sentence is the authors' text from the arXiv listing abstract. Labels and probabilities are the outputs of classifier-1.13.0 on the question bank abstract-p1-20260923, exactly as returned. No sentence in this record was written by a language model.","schema":"pith.compiled-record/v0","status":"compiled","abstract":{"source":"arXiv listing abstract v1","sentences":[{"i":0,"p":0.85,"id":"s0","own":false,"role":"background","text":"Recursive self-improvement (RSI) of a model on non-verifiable tasks, such as open-ended research, faces a supervision bottleneck when its outputs exceed what even human experts can reliably assess, leaving the model itself (optimizee) as the best available optimizer and evaluator.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.07,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.29,"role_probabilities":{"gap":0.15,"method":0.0,"result":0.0,"background":0.85,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":1,"p":0.67,"id":"s1","own":false,"role":"gap","text":"However, a single model instance struggles to critique and improve its own complex reasoning under this homogeneous loop.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.14,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.67,"method":0.0,"result":0.0,"background":0.29,"limitation":0.04,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":2,"p":0.84,"id":"s2","own":true,"role":"contribution","text":"To address this, we propose Multi-Agent Self-Supervision (MASS), an RSI method that alternates between evolutionary workflow optimization and supervised fine-tuning on self-generated trajectories.","hedge":{"p":0.96,"label":"asserted"},"own_p":0.68,"main_claim_p":0.13,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.0,"method":0.16,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.84,"interpretation":0.0}},{"i":3,"p":0.96,"id":"s3","own":true,"role":"method","text":"Guided by early findings that multi-agent topologies excel at complex reasoning, MASS prompts a single base model to iteratively propose, execute, and self-evaluate multi-agent workflows.","hedge":{"p":0.91,"label":"asserted"},"own_p":0.58,"main_claim_p":0.06,"quantitative":false,"quantitative_p":0.11,"role_probabilities":{"gap":0.0,"method":0.96,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.04,"interpretation":0.0}},{"i":4,"p":1.0,"id":"s4","own":true,"role":"method","text":"Through an evolutionary search constrained by structural guardrails, the model optimizes these computational-graph-like orchestrations, discovering the most effective distinct roles and information routing for a given task.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.82,"main_claim_p":0.07,"quantitative":false,"quantitative_p":0.22,"role_probabilities":{"gap":0.0,"method":1.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":5,"p":1.0,"id":"s5","own":true,"role":"result","text":"Over two MASS cycles with Qwen3.6-27B, the model achieves 1.2-1.6x higher performance per output tokens on four open-ended public benchmarks.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.96,"main_claim_p":0.21,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":6,"p":0.93,"id":"s6","own":true,"role":"interpretation","text":"Because the improved model subsequently acts as a better optimizer and evaluator, this alternating framework enables a continuous, recursive bootstrapping of the model's capabilities.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.62,"main_claim_p":0.29,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.0,"method":0.04,"result":0.01,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.02,"interpretation":0.93}},{"i":7,"p":1.0,"id":"s7","own":true,"role":"result","text":"Moreover, multi-agent traces are also more training-efficient: a student trained on them outperforms a single-agent student trained on 1.4x more training tokens.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.89,"main_claim_p":0.08,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":8,"p":1.0,"id":"s8","own":true,"role":"interpretation","text":"These findings suggest that jointly learning orchestration and bounded subagent execution from multi-agent trajectories can provide an effective signal for RSI.","hedge":{"p":1.0,"label":"hedged_possibility"},"own_p":0.82,"main_claim_p":0.71,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":1.0}}]},"compiled":{"main_claim":{"p":0.99,"text":"These findings suggest that jointly learning orchestration and bounded subagent execution from multi-agent trajectories can provide an effective signal for RSI.","location":"abstract s8","hedge_strength":{"p":1.0,"label":"hedged possibility"}},"provenance":{"route":{"via":"TypeSafe API","model":"classifier-1.13.0","endpoint":"https://api.typesafe.ai/v1/systemone","provider":"TypeSafe"},"packing":"one request per paper","requests":1,"input_tokens":7084,"cost_usd_reported":0.0},"compiled_at":"2026-10-09","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.04,"states_limitation":0.45,"human_participants":0.06,"controlled_comparison":0.82,"randomized_assignment":0.02,"states_quantitative_result":0.98,"compares_with_prior_results":0.77,"mentions_code_or_data_availability":0.03},"evidential_basis":{"p":0.94,"value":"observation or experiment","probabilities":{"argument":0.0,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.0,"observation_or_experiment":0.94,"simulation_or_computation":0.06}},"contribution_type":{"p":1.0,"value":"method or tool","probabilities":{"review":0.0,"method_or_tool":1.0,"empirical_study":0.0,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":1.0,"text":"To address this, we propose Multi-Agent Self-Supervision (MASS), an RSI method that alternates between evolutionary workflow optimization and supervised fine-tuning on self-generated trajectories.","location":"abstract s2"}},"coverage":{"tier":"abstract","levels_run":["P abstract"],"not_examined":["full text","figures and tables","references and citing sentences","relations to other papers"],"sources_read":["arXiv listing abstract v1"],"levels_not_run":["S sections","T body sentences","C citations","R cross-paper relations"],"sentences_examined":9,"sentences_in_abstract":9},"measured_accuracy":{"url":"https://pith.science/accuracy","measured_on":"2026-10-02","method":"two independent model families labelled a fresh random sample blind; the figure is how often the published label matches their consensus","sentence_role_agreement":0.92,"main_claim_agreement":0.908,"typed_citation_agreement":0.933,"typed_citation_agreement_95_interval":[0.917,0.945],"review_comments_accurate":0.758,"review_comments_accurate_95_interval":[0.684,0.82],"review_caught_admitted_error_found_or_related":0.448,"review_comments_judged":{"judged_by":["gpt-5.6-luna","grok-4.7"],"judged_against":"Pith's own text copy of each paper: text extracted from the PDF, the first 90,000 characters","measured_on":"2026-10-04","accurate":113,"decided":149,"sampled_reviews":200,"split_by_lane":false,"by_lane":"https://pith.science/accuracy#by-lane"}}}